Content
86%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
An excellent, disciplined instruction-only skill body: a concrete per-PR report template backed by unusually thoughtful decision rules and an injected untrusted-input safety policy, with zero wasted tokens. Its one substantive gap is that data-gathering commands and the meaning of the suggested "minimal-fix" action are unspecified, which keeps actionability and workflow clarity just short of exemplary.
Suggestions
Add the concrete commands used to gather each field, e.g., `gh pr view --json checks,reviews,mergeable,state` or `gh pr checks <N>`, so the reporting workflow is fully executable end to end.
Define what the "minimal-fix" loop action entails (scope, limits, when to fall back to escalate-human) since it is offered as a suggested action but never specified.
Make the per-PR sequence explicit (fetch checks/reviews/mergeability → apply rules → emit template) so the evaluation order is stated rather than implied by the template.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence: a compact output template plus tight, non-obvious rules ("Zero checks... means absent/unknown, not passing"; "mergeable... only means Git found no conflict") with no padding or explanation of concepts Claude already knows. Every rule encodes a real decision norm, matching the every-token-earns-its-place anchor. | 5 / 5 |
Actionability | The per-PR output template is copy-paste ready with enumerated states and thresholded rules ("idle >4 days", "high-risk labels → escalate-human"), but no commands for gathering the data (e.g., gh pr view --json, gh pr checks) are given. Concrete guidance with a real input-gathering gap matches the mostly-executable anchor 4 rather than the fully-executable anchor 5. | 4 / 5 |
Workflow Clarity | Template fields paired with decision rules form a clear procedure with explicit checkpoints (policy must be established or escalate; administrative statuses cannot substitute for test results). It stays below anchor 5 because the sequence is only implicit in template order and the suggested "minimal-fix" action is never defined, leaving a minor ambiguity. | 4 / 5 |
Progressive Disclosure | The body is under 50 lines, self-contained with no external references needed, and organized into clear sections (Per-PR Output, Rules, Untrusted input), satisfying the rubric's simple-skill criterion for a top score. The only path mention is a maintainer-facing provenance comment, not model-facing navigation, so navigation is unimpaired. | 5 / 5 |
Total | 18 / 20 Passed |